Approaching Persistent Problems in Health IT Through Science
HIMSS 2019 – Orlando, Florida
• Teresa Zayas Cabán, PhD, Chief Scientist
Office of the National Coordinator for Health Information Technology
• Ricky Sahu, CEO, 1upHealth
• Kristen Miller, DrPH, CPPS, Scientific Director
National Center for Human Factors in Healthcare at MedStar Health
Agenda
• Teresa Zayas Cabán, ONC
» ONC Scientific Initiatives
• Ricky Sahu, 1upHealth
» Pop Health on FLAT FHIR: A SMART Approach to Universal Healthcare Reporting
• Kristen Miller, MedStar Health National Center for Human Factors in
Healthcare
» Mobilizing a Million Hearts: Leveraging Health IT Architecture to Advance Clinical
Knowledge and Care Coordination
2
At the Intersection Between Research and Care Delivery
• Develop and evaluate ONC’s scientific efforts and activities
• Recommend scientific policy to the National Coordinator
• Promoting activities that spur innovation, support patient-centered
outcomes research, and advance precision medicine
3
https://www.healthit.gov/topic/scientific-initiatives
Patient-Centered Outcomes Research (PCOR) Projects
• Goal: To develop the policy,
standards, and services
necessary to expand the data
infrastructure for patient-
centered outcomes research
• Current Projects:
» Coordinated Registry
Network for Women’s
Health Technologies
» Common Data Model
Harmonization
» Patient-Reported Outcomes
through Health IT
4
https://www.healthit.gov/pcor
Sync for Science
• Goal: Promote patient-mediated
access to data; establish structures
that facilitate data sharing to the PMI
cohort
• National collaboration among EHR
developers, NIH, ONC, and Harvard
Medical School’s Department of
Biomedical Informatics
http://syncfor.science/
6
Patients Sharing EHR Data
Sync for Genes
• Goal: To standardize the sharing of
genomic information between
laboratories, providers, patients, and
researchers
• Current Phase 2: Integrating Genomic
Data
» Support integration of genomics
information and clinical information at
the point of care
7
https://www.healthit.gov/sites/default/files/sync_for
_genes_report_november_2017.pdf
Advancing Standards for Precision Medicine
• Goal: To advance
standards development
and use for sensors/
wearable and social
determinants of health
data
• Collaboration with
Open mHealth
8
Policy and Development Agenda
• Goal: To advance the nation’s
health IT infrastructure over the
next 3 to 5 years in support of
advancements in biomedical
and health services research
A health information ecosystem
where research happens faster,
better, easier, and improves
outcomes
10
Leading Edge Acceleration Projects (LEAP) in Health IT
• Goals:
» Address challenges to the
development, use, and/or
advancement of well-designed,
interoperable health IT
» Develop solutions and advances to
further a new generation of health
IT development
• Areas of Interest:
1. Population-level data focused APIs
2. Advancing clinical knowledge at the
point of care
11
1. Children’s Hospital Corporation
• Pop Health on FLAT FHIR:
A SMART Approach to
Universal Healthcare
Reporting
2. MedStar Health Research
Institute
• Mobilizing a Million Hearts:
Leveraging Health IT
Architecture to Advance
Clinical Knowledge and Care
Coordination
Flat FHIR LEAP Award
@smarthealthit
@bos_chip
FLAT FHIR LEAP GRANT TEAM
Ken Mandl, MD PHD, from Boston Children’s
Hospital, is the Director of the Computational
Health Informatics Program (CHIP) team which is
behind SMART Health IT.
Ricky Sahu, CEO of 1upHealth, leads a team
building FHIR APIs and patient mediated
exchange of EHR data between 100s of apps and
health systems. Ricky Sahu is working with CHIP to
produce the open source LEAP products.
ELECTRONIC HEALTH
National investment promoting
adoption of EHRs assumed basic access
to support machine readable access of
the data, but more work is needed.
The goal is
“push button
population health”
The birth of flat FHIR
One year ago, in a meeting of government,
payors, and technology companies, requirements
for bulk data export were developed.
Over the past year, a specification and reference
implmentation have been produced for bulk data
export in a FHIR format
BULK DATA / FLAT FHIR
• Using NDJSON format
• Able to transmit data in bulk
• Adding a security layer for
trusted servers
LEAP GRANT
Don Rucker and HHS identified an
excellent initial use case:
REDUCE Payor to Provider Reporting
Burden
PROBLEM
Health systems in risk-bearing
contracts, bear large administrative
costs to meet often nuanced and
redundant reporting requirements to
payor orgs including CMS
• No manual data entry for reporting
• Standardized outputs in a machine readable
format
• Data are transmitted or accepts interrogation by
report requestor
• Payors or regulators provide the reporting
queries
THE PROMISE
Expanding the scope, scale, and utility of
population-level data-focused APIs
We are developing and testing at production scale,
an open source, reference population health app
which for use between payers and hospitals.
LEAP GRANT GOAL
Specific Objectives
1.Reducing provider burdens associated with
reporting through this technology;
2.Investigating and assessing trade-offs
associated with various big data formats; and
3.Technical and legal/policy challenges to the
scope and scale of FHIR-based APIs for these
purposes.
Architecture
Flat
FHIR
Files
Flat
FHIR
Files
Flat
FHIR
Files
BCH Cerner
Millennium
EHRs and data
warehouses become a
distributed database,
with federated query
and analytics across
healthcare systems
(Years 3-5)
BCH Enterprise
FHIR Server
(Claims + FHIR
Data)
SMART
Health Bulk
Data
Reference
Server
SMART
PopHealth
Application
Analytics
On Flat FHIR
FHIR Bulk
Data Export
API
Health
System
#2 with
Bulk Data
API
Flat
FHIR
Files
oAUTH2
PKI
Jason Web
Tokens
Payor Claims
Data FHIR server
with Bulk Data
API (Years 3-5)
Health
System
#3 with
Bulk Data
API
• FLAT FHIR Files - NDJSON files w security
• Analytics Optimized Files / System - Avro,
Parquet, Hadoop, or NDJSON
• Big data, distributed analytics engine - HIVE,
Spark, Impala, Presto
• SMART PopHealth Client Application – custom
queries against analytics engine & displays result
SMART PopHealth
Data Flow
Existing Technology
(Provider organization)
Flat FHIR Server
(Provider organization)
SMART Pop Health App
Flat FHIR Client
(Payor organization)
EHR FHIR
API Server
Analytics
Query Engine
Smart Pop
Health Client
App
Flat FHIR
(Bulk Data)
Server
Data Flow Detail
Existing Technology
(Provider organization)
Flat FHIR Server
(Provider organization)
SMART Pop Health App
Flat FHIR Client
(Payor organization)
FHIR API
Server
FHIR Client to
read all
historical and
recent data
Flat FHIR
Reader
Client
Smart Pop
Health Client
App
Analytics
Query Engine
Frontend
viewer for
reports and
metrics
FHIR
Queries
Flat
FHIR
Files as
NDJSON
Avro / NDJSON
/ Parquet files
FHIR
Resourc
es
Flat
FHIR
Queries
SQL
Queries
Tabular
data
Tabular
data
Flat FHIR
(Bulk Data)
Server
FHIR
Data into
cache
Database
• Tech
o How does it scale on multiple machines?
o Can reporting requirements be met on FHIR?
• Legal / Policy
o Are health systems ok with divulging such
granular data?
o How do we manage opt-outs for data sharing?
Technical, Legal, & Policy considerations
28
Mobilizing a Million Hearts:
Leveraging Health IT Architecture to
Advance Clinical Knowledge and Care
Coordination
Kristen Miller, DrPH, CPPS
Scientific Director
National Center for Human Factors in Healthcare
MedStar Institute for Innovation, MedStar Health
Associate Professor of Emergency Medicine, Georgetown University
Objectives
• ONC Leap addresses well-documented and fast emerging challenges
inhibiting the development, use, and/or advancement of well-
designed interoperable health information technology.
• The purpose of our project is to:
1. Support evidence-based clinical cognitive support that prompts
clinical management and promotes preventative care.
2. Serve as proof-of-concept to transform risk calculators into
active surveillance tools leading to guideline based workflow
support through SMART on FHIR technology.
3. Leverage the technology to facilitate communication and
coordination within providers, and between providers and patients
as engaged members of their care with reduced clinical burden.
Background: Million Hearts
• Cardiovascular disease remains the
leading cause of death in the US.
• The American Heart Association and
American College of Cardiology
recommend use of the Atherosclerotic
Cardiovascular Disease (ASCVD) risk
estimator: evaluates 10-year and
lifetime risk for ASCVD.
• Variables include:
» Age and Race
» Cholesterol levels (HDL, LDL)
» Blood pressure
» Use of statin therapy
» Antihypertensive medication
» Use of aspirin therapy
» Smoking status
» Diabetes status
Background: Million Hearts Optimization
Our research addresses the following:
1. Optimizing current health IT tools: remove
the burden of active surveillance, push
relevant data to the clinician.
2. Reducing time required to integrate
clinical guidelines at the point of care.
3. Developing solutions that are not product
centric – our solution sits outside of the
EHR and does not rely on the vendor to
support modifications.
4. Developing solutions that integrate into
clinician and patient workflow.
5. Developing scalable solutions that change
the way we think about patient data and
decision support (multi-layered support
and visualizations).
Specific Aims
Specific Aims
Specific Aims
Workflow Analysis
Methods:
• Stakeholder Interviews
» 6 Cardiologists, 7 Primary Care Physicians, 4 Care Navigators
» In Progress: System-wide end-user survey
• Clinical observations
» 30 hours = 34 observed patient visits
» In Progress: Citrix SmartAuditor EHR use review
• Data Analysis
» “Work-as-imagined” versus “Work-as-done”
Stakeholder Interviews: Goals
• Develop a detailed understanding of current state use of cardiac risk
calculators in both primary care and cardiology settings – by both physicians
and care navigators.
• Understand the barriers to communicating risk to patients.
• Understand how the calculators may be used to facilitate care coordination.
Stakeholder Interviews: Results
3 Main Uses
• To educate patients about
managing cardiovascular risk.
• To aide in clinical decision
making about whether or not to
prescribe a statin.
• To identify, in borderline cases,
whether or not a patient is at risk
of cardiovascular disease.
Cardiology
Participants
(n = 6)
Primary Care
Participants
(n = 7)
N (%) N (%)
Medical Specialty
Cardiology 6 (100%)
Internal Medicine 5 (71%)
Family Medicine 2 (29%)
Years in Practice (including residency)
< 5 1 (16%) 2 (29%)
5-10 2 (32%) 1 (14%)
11-15 1 (16%) 1 (14%)
16-20 1 (16%) 1 (14%)
> 20 1 (16%) 2 (29%)
Setting
Out-patient only 3 (43%)
Both in-patient and outpatient 6 (100%) 4 (57%)
Electronic Health Record Vendor (current)
Cerner
4 (66%)
7 (100%)
Epic 2 (33%)
Current ASCVD Risk Calculator Use (self
-reported)
Yes 6 (100%) 7 (100%)
Clinical Observations: Goals
• Develop a detailed understanding
of current physician workflow in
both primary care and cardiology
settings.
• Develop a detailed understanding
of how and when ASCVD risk
factors are addressed over the
course of a typical patient exam.
• Develop detailed process maps
for how ASCVD risk calculators
are utilized.
Clinical Observations: Results
• Complex cardiology visit.
• Complex primary care practice visit.
• Simple cardiology visit.
• Simple primary care practice visit.
• Use of ASCVD risk calculator during visit.
Simple Cardiology Visit Process Map
Complex Primary Care Visit Process Map
Use of ASCVD Risk Calculator During Visit Process Map
Specific Aims
Specific Aims
Application Development
Methods:
• Use case development to inform technical specifications
» Matching functions to features to technical requirements
• User interface design
» Design workshops
Use Case
Development
1. Risk Assessment:
» Physician assesses new patient
» Physician reassess patient’s risk
» Notification of calculator specific incoming
values
» Notification to physician for patient’s risk
level change
2. Clinical Decision Making:
» Medications: decision to prescribe statins
» Diagnostics: decision to run labs
3. Patient Education:
» Physician shares current risk with patient
» Physician demonstrates improved risk given
behavior/medication modification
4. Care Coordination:
» Shared decision making across providers
» Risk awareness for care navigators
Functions to Features (sample)
Function
Feature
Technical Requirements
Calculate ASCVD risk (with
minimal clinical burden)
Autopopulate
values into the
ASCVD risk calculator
FHIR
Ensure ASCVD risk is complete
and up
-to-date
Highlight missing/outdated
information
FHIR and UX
Recalculate score
Refresh ASCVD risk when any
new calculator related value is
entered
CDS hooks
Cerner SmartZone
Evaluate risk score trends
Display risk scores (current and
previously saved)
FHIR
Alert physician to change in risk
score
Trigger new calculation when
patient undergoes outside
procedure (e.g., surgery, ED)
that changes risk variables given
set parameter (e.g., change in
risk level)
Cerner Message Center API
FHIR
CDS Hooks
Population health approach to
evaluate risk for entire patient
panel
Display patient list, organized by
associated risk
Subscriptions (bulk FHIR)
UX
User Interface Design
User Interface Design
Next Steps
